Dr. Jacob Feldman talked about rules and the problem of some rules that get violated because they were over-constrained. An example solution to this is to introduce “soft” constraints and an associated cost to the violation.

He identifies 2 types of rules:

• Hard rules: that must be satisfied
• Soft rules: that we would like to be satisfied

The idea is to use the experience of Constraint Programming (CP) to help with these “over-constrained problems”.

He gave us an example of constraint resolution for a problem of a map of countries that need to be coloured with different colours and that no two neighbouring countries can have the same color. With the 6 countries and the 4 colours, there is a solution.

Now if you take this same problem with a limit of 2 colours, there is no solution, but we can “soften” the constraints and to try to find a solution to the same problem.

You can apply this strategy to business rules by adding some constraints on violation of other constraints. These “meta-constraints” are themselves business rules. He also thinks that because this can be implemented in business rules, the business people can actually control the constraints themselves.

The application of this type of hard and soft rules requires a CP expert to be added to the list of people involved in the implementation.

He then showed an example of the well know “Miss Manners” problem with hard constraints. He then used some of the typical data and changed the gender of one of the “guests” and the problem became unsolvable (in a reasonable amount of time). By “softening” some of the constraints it is possible to find solutions.

He concluded that it is possible to do some of this type of work using “Greedy Java” so Java is a definitive competitor to business rules and constraint programming. Obviously the rules are not declarative, the solution is not optimized, etc.

Very interesting talk if you have the type of problems to resolve that requires CP.